It's just a tiny tease and raises a whole host of questions, but in a recent online interview Mark Papermaster was asked what the year has in store for AMD. The CTO's answer was, of course, focused solely on AI, but he also mentioned a curious little tidbit: its gaming devices are set to receive AI-based upscaling and 2024 will likely be the year we'll see it.
The discussion with YouTube channel No Priors (above Videocardz) is mainly about AMD's position and thoughts on machine learning, cloud computing and other major sectors. It's a good conversation between the hosts and Papermaster, if a bit dry if you're not interested in these particular topics.
However, the CTO's closing remarks on AMD's 2024 plans are certainly of interest to gamers, especially thanks to the brief commentary on how the chip giant is bringing AI upscaling to its gaming devices.
“Well, this is a big year for us because we have spent so many years developing our hardware and software capabilities for AI,” Papermaster begins. “We just completed the AI that enables our entire portfolio: cloud, edge, PCs, embedded devices, our gaming devices. We are enabling upscaling of our gaming devices using AI and 2024 is really a big launch year for us.”
Currently, AI upscaling is reserved for Intel and Nvidia. Both companies use a two-step upscaling process, where the first step is to run an image scaler via shaders to bring the resolution of the rendered frame up to the level of the monitor. The second step involves applying a neural network to clean up the image and remove any noise caused by the upscaling process.
The mathematical operations required for this can be performed on shaders, but in the case of DLSS and the Intel Arc-only version of XeSS, they are processed by dedicated matrix units in the GPU to achieve the best performance and visual results. The graphics chips used in AMD's Radeon models perform all calculations via the Compute Units (CU), which is why FSR does not require an AI step to clean up the images.
However, one interpretation of Papermaster's comment is that the next graphics architecture, RDNA 4, could have a discrete matrix unit in each CU and if so, this could be used for AI upscaling. An alternative view is that AMD has developed a relatively lightweight neural network that is managed by the CUs and can be used as an additional step in the FSR pipeline to improve visual quality.
As things stand, FSR and DLSS perform about the same, but the latter has the edge when it comes to graphical fidelity. Depending on the game it can be significantly better, but it is a very subjective thing during gameplay.
If AI upscaling in FSR is implemented so that any GPU can do it (as long as it supports the relevant matrix instructions), one can expect an improvement in quality, but performance will likely decrease.
It's the way Papermaster says “gaming devices” that makes me think AMD's AI upscaling is most likely happening this way. Because if it was just RDNA 4, you'd expect him to mention it, or at least talk specifically about GPUs.
However, the term “gaming devices” includes anything that supports FSR, including current Radeon cards as well as the APUs in handheld PCs and consoles. If it turns out that dedicated matrix units are required, as is the case with DLSS, it will only be supported on AMD's next GPU architecture and is nothing special for the Xbox Series X or PlayStation 5.
RDNA 4, regardless of its AI capabilities, will compete against Intel's Battlemage and Nvidia's Blackwell sometime in 2024, and these companies already have AI-powered upscaling in place.
AMD is fashionably late to the party, but could steal the spotlight if it can offer better upscaling that all GPUs can take advantage of. Papermaster is certainly right about one thing: 2024 will be a big year for AI and GPUs.
